powerpc/eeh: Enable PCI_COMMAND_MASTER for PCI bridges
authorGavin Shan <shangw@linux.vnet.ibm.com>
Tue, 12 Nov 2013 06:49:21 +0000 (14:49 +0800)
committerBenjamin Herrenschmidt <benh@kernel.crashing.org>
Wed, 20 Nov 2013 23:33:35 +0000 (10:33 +1100)
On PHB3, we will fail to fetch IODA tables without PCI_COMMAND_MASTER
on PCI bridges. According to one experiment I had, the MSIx interrupts
didn't raise from the adapter without the bit applied to all upstream
PCI bridges including root port of the adapter. The patch forces to
have that bit enabled accordingly.

Signed-off-by: Gavin Shan <shangw@linux.vnet.ibm.com>
CC: <stable@vger.kernel.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
arch/powerpc/kernel/eeh.c

index 67130206534717b285fb50c77d38e4557f053eae..4bd687d5e7aa3f80e1b33912b5451bd34a3187b1 100644 (file)
@@ -686,6 +686,15 @@ void eeh_save_bars(struct eeh_dev *edev)
 
        for (i = 0; i < 16; i++)
                eeh_ops->read_config(dn, i * 4, 4, &edev->config_space[i]);
+
+       /*
+        * For PCI bridges including root port, we need enable bus
+        * master explicitly. Otherwise, it can't fetch IODA table
+        * entries correctly. So we cache the bit in advance so that
+        * we can restore it after reset, either PHB range or PE range.
+        */
+       if (edev->mode & EEH_DEV_BRIDGE)
+               edev->config_space[1] |= PCI_COMMAND_MASTER;
 }
 
 /**